Textbook Notes (270,000)
CA (160,000)
UTSC (20,000)
Psychology (10,000)
PSYB51H3 (300)
Chapter 4

Chapter 4 Notes

Course Code
Matthias Niemeier

This preview shows pages 1-2. to view the full 7 pages of the document.
Chapter 4: Perceiving and Recognizing Objects
-e.g. a house, simple cortical cells only see high contrast lines
-only sees parts, dont actually see house. Just lines.
-Middle (Midlevel) vision: Stage in seeing, after basic features are interpreted (early
vision). But before object recognition and scene understanding.
-combines features into objects
-e.g. how we a snowmans edges dont belong to the house
-recognition must involve matching what we see to memory
Middle Vision
-goal is to sort elements in scene into groups that we see as objects
Finding Edges
-cant assume lines that touch each other are objects b/c they overlap
-we fill in edges even when colours may be the same and there are gaps
-b/c of random in lighting
-illusory contour: When we see edges that arent actually connected
-e.g. fig. 4.5
-structuralism: Complex perception can be understood by analyzing components
-i.e. perceptions are made of atoms from sensation (colour, orientation)
-see perception as built up from local sensations of atoms
-illusory contour disproves b/c gaps (cant have atoms) can be seen as having edges
-Gestalt: Whole is > sum of sensory parts
-german means form, everything has form, but we cannot always see it
-Gestalt grouping rules: rules for which elements are grouped from raw retinal image
-lines of sim. Orientation are seen as part of same contour
-parallel lines with gap btw support each other, make it easier to see lines between
them-e.g. fig 4.6
-even more so if the lines form a closed shape
-good continuation: gestalt grouping rule, elements group together if they seem to lie on
same contour
-e.g. fig 4.7
-we see lines that continue in one direction
-group edges that have seem orientation
-perception is like committee, we take different rules, principles and guesses and come to a

Only pages 1-2 are available for preview. Some parts have been intentionally blurred.

-one make overtake another, e.g. colour tells us contour as opposed to good continuation
-visual system wonders why an edge stops
-it may guess that something is blocking it
-when we combine this with good continuation, we get the illusory contour
Texture Segmentation and Grouping
-edge finding isnt very useful for texture differentiation
-texture differentiation: separating image into places with the same textures
e.g. fig 4.9 larger polygons are separated from smaller ones
-based on two principles:
Similarity: images that look the same are grouped together
e.g. similar colour, size, orientation and form
-combinations dont work well. i.e. difficult to tell diff between a combo of
features in image
Proximity: things that are closer are more likely to be grouped
-parallelism: parallel images likely grouped
-symmetry: symetrial images likely grouped
-common region: images likely grouped together if they look like part of a larger region
e.g. fig 4.12 line 2
-connectedness: images likely grouped together if they are connected
-camouflage is getting features to group with environment when they shouldnt be
Perceptual Committees Revisited
-middle vision is like collection of specialists for different inputs from low level vision
-goal is to have single image out of these opinions
-committees are like connected neurons that take inputs and send them elsewhere to make
a image
-decisions made by perceptual committees dont have to be final
-ambiguous feature: image that gives 2+ interpretations
-e.g. necker cube: cube that can be viewed in two ways fig 4.15
-accidental viewpoint: viewpoint that makes shapes look regular even though they
maybe oriented and placed differently
-e.g. fig 4.17
You're Reading a Preview

Unlock to view full version